Overclocking RAM, Gear 1 vs Gear 2

[Pets4Ever]

I'm on AM5, 32GB DDR5-6000 CL28. If I were to overclock the ram to 6200, would that push my memory controller into Gear 2?

 

[EarthDog]

Likely not with 2x16GB sticks. Depends on the board. My 64GB kit does on an Asus board... not sure on that Tomahawk.

Try it and see...report back!

 

[Chrispy_]

I'm on AM5, 32GB DDR5-6000 CL28. If I were to overclock the ram to 6200, would that push my memory controller into Gear 2?

You're using an X3D chip, stop trying to get 3% more out of your RAM because the X3D chip doesn't care in the slightest, doesn't need more bandwidth, and doesn't need lower latency.

The only thing that comes from pushing your memory controller to the limits on an X3D chip is increased idle power draw and potential instability that doesn't show up very easily thanks to DDR5's inherent single-bit ECC.

Essentially, you're gaining almost nothing, definitely worsening your power consumption and temperatures, and pushing your memory controller from further from the guaranteed speeds its rated for. It's like tuning an engine to within an inch of its life when all it actually needs to do it drive two miles to the shops in heavy, slow-moving traffic. Look at articles that show DDR5 scaling on 7800X3D and 9800X3D. It's damn-near pointless so the risk-reward is basically all risk, zero reward; Tune literally anything else in your system first.

 

[ir_cow]

I'm on AM5, 32GB DDR5-6000 CL28. If I were to overclock the ram to 6200, would that push my memory controller into Gear 2?

It could, but you can always enforce ULCK/Mem/1 ratio in the BIOS. Will it boot? Probably, will it be stable, probably.

2x single rank 6400-6600 is where you start to have problems. This is of course excluding FCLK and the optimal ratio for that.

 

[Pets4Ever]

You're using an X3D chip, stop trying to get 3% more out of your RAM because the X3D chip doesn't care in the slightest, doesn't need more bandwidth, and doesn't need lower latency.

The only thing that comes from pushing your memory controller to the limits on an X3D chip is increased idle power draw and potential instability that doesn't show up very easily thanks to DDR5's inherent single-bit ECC.

Essentially, you're gaining almost nothing, definitely worsening your power consumption and temperatures, and pushing your memory controller from further from the guaranteed speeds its rated for. It's like tuning an engine to within an inch of its life when all it actually needs to do it drive two miles to the shops in heavy, slow-moving traffic. Look at articles that show DDR5 scaling on 7800X3D and 9800X3D. It's damn-near pointless so the risk-reward is basically all risk, zero reward; Tune literally anything else in your system first.

I guess I should go back to my 12700k and gtx 1070. You know, to save electricity.

 

[Chrispy_]

Even with a very very pricey DDR5-8400 kit and locking your FCLK and UCLK to 2100 in sync, it's still less than 1% faster than DDR5-6000

Power consumption of your CPU isn't about saving electricity, it's about PPT power tracking and it affects your boost clocks, which has a much bigger impact on performance than RAM tuning. Don't get me wrong, the performance gains from a couple of Watts more power budget are also insignificant, but they're bigger than the absolutely miniscule gains of adding 3% RAM bandwidth to an X3D chip.

Feel free to try it at 6200. Make sure you take an average of repeatable benchmarks of 6000 and 6200, chances are good you'll see no change whatsoever, and with most BIOSes, the PBO+ limits for a 9800X3D means you might even be able to measure a small performance regression from moving to DDR5-6200 Gear1. I personally think it'll be within the margin of testing error but that's my guess if you were able to get enough consistent tests averaged out.

 

[Pets4Ever]

Even with a very very pricey DDR5-8400 kit and locking your FCLK and UCLK to 2100 in sync, it's still less than 1% faster than DDR5-6000

Power consumption of your CPU isn't about saving electricity, it's about PPT power tracking and it affects your boost clocks, which has a much bigger impact on performance than RAM tuning. Don't get me wrong, the performance gains from a couple of Watts more power budget are also insignificant, but they're bigger than the absolutely miniscule gains of adding 3% RAM bandwidth to an X3D chip.

I'd like to see a study on 1% and .01% lows related to ram speed on the 9800X3D. It's hard to believe there is no benefit.

I would assume that faster data-rate should mean faster loading times, right? How can the 3D cache of 96mb can make much difference when you're loading, say, 8 gigs into ram all at once?

 

[ir_cow]

Ahh the debate continues. Check out this thread and my not so popular opinion.

Post in thread 'Need help overclocking my 9800X3D'

Jun 4, 2025

You may want to learn a little more about memory overclocking and tuning on AM5.

You are missing the point im making. If your GPU bound, there won't be any extra fps to be had.

It's true I cannot test ever game and setup scenario, but I can generalize that the impact in minimal. I much rather take that $100 saved and buy a better video card.

If we are going to start talk about overclocking, thats a whole different topic and ways to achieve higher fps. Assuming there is no GPU bound limitations, memory latency will be slightly impactful, but higher CPU clocks increase to the FLCK...

• ir_cow

 

[Chrispy_]

I'd like to see a study on 1% and .01% lows related to ram speed on the 9800X3D. It's hard to believe there is no benefit.

I would assume that faster data-rate should mean faster loading times, right? How can the 3D cache of 96mb can make much difference when you're loading, say, 8 gigs into ram all at once?

Well you're in luck, friend - the internet is full of reviews of the 9800X3D, many of those reviews test DDR5 performance scaling and all of them prove there is no significant benefit. You'll probably see DDR5 6000, 6200, and 8000 all swapping podium places at random, because the measurement error is greater than the actual performance difference.

If you were running a 9700X I would 100% recommend pushing your memory clock to the stable limit because those CPUs do get measurable benefits from lower latency and more bandwidth, they also don't have the PPT power/voltage control restriction and cooling difficulties that X3D chips have. Yes, the 9800X3D is easier to cool than the 7800X3D because the 3D cache is underneath the cores, but they still run hotter than a similarly clocked and volted 9700X, and they still don't like high voltages or aggressive PBO limits.

 

[freeagent]

I dont know about in game, I would imagine it would benefit. I have been looking into DDR5 OC, and from what I have seen, outside of games, X3D does see a substantial boost from memory tweaks.

 

[Pets4Ever]

Chrispy_, I'm setting you to Ignore. You're rude and condescending and I don't need that in my life. It's entirely possible to distribute information without being the way you are. Don't bother responding to this because I won't see it. If you can't figure out what's wrong with your posts, it's not my job to teach you.

Ahh the debate continues. Check out this thread and my not so popular opinion.

Post in thread 'Need help overclocking my 9800X3D'

Jun 4, 2025

You may want to learn a little more about memory overclocking and tuning on AM5.

You are missing the point im making. If your GPU bound, there won't be any extra fps to be had.

It's true I cannot test ever game and setup scenario, but I can generalize that the impact in minimal. I much rather take that $100 saved and buy a better video card.

If we are going to start talk about overclocking, thats a whole different topic and ways to achieve higher fps. Assuming there is no GPU bound limitations, memory latency will be slightly impactful, but higher CPU clocks increase to the FLCK...

• ir_cow

In modern games with horrible optimization, isn't it possible to have spikes of incredibly high CPU activity that cause a split-second "bottleneck" while the rest of the time it is essentially GPU limited?

 

[ir_cow]

Chrispy_, I'm setting you to Ignore. You're rude and condescending and I don't need that in my life. It's entirely possible to distribute information without being the way you are. Don't bother responding to this because I won't see it. If you can't figure out what's wrong with your posts, it's not my job to teach you.

If you think @Chrispy_ is rude, I fear you aren't going to like the rest of the forum members, Reddit or the Internet.

Not much more say on the subject. Lot of debates, lot of proof that is really doesn't matter in games as much as people make it out to be. The scenario in which ram latency matters in games for the X3D is not what I would consider "normal gaming" settings.

 

[Pets4Ever]

If you think @Chrispy_ is rude, I fear you aren't going to like the rest of the forum members, Reddit or the Internet.

Not much more say on the subject. Lot of debates, lot of proof that is really doesn't matter in games as much as people make it out to be. The scenario in which ram latency matters in games for the X3D is not what I would consider "normal gaming" settings.

What about my question regarding load times? How can 96mb of cache matter when loading 8gb into ram at once?

 

[Chrispy_]

I dont know about in game, I would imagine it would benefit. I have been looking into DDR5 OC, and from what I have seen, outside of games, X3D does see a substantial boost from memory tweaks.

Where are you looking and what workloads, out of interest? Rendering and encoding are the sort of non-gaming workloads that take long enough to merit an improvement, and neither of those seem to care about memory speed (within reason) on a 9800X3D according to TPU's article:

DDR5 Memory Performance Scaling with AMD Zen 5

Our AMD Zen 5 Memory Scaling Review examines six different DDR5 memory speeds, including DDR5-8000, DDR5-6400 1:1, DDR5-6400 CL28 and others. We looked into how these speeds impact performance in gaming at four resolutions and a wide range of application workloads.

www.techpowerup.com

I didn't check every single result, and TPUs test can't possibly cover every application out there - but the benefits across a wide range of applications do seem to indicate that the X3D chips really don't care much about RAM speed beyond about 6000 which is probably why everyone from here to the moon parrots "DDR5-6000 is the sweet spot" still.

Hmm, on further reading Adobe Premiere/After Effects show some significant scaling though it looks like W1zzard went out of his way to find one of the few non-GPU-accelerated operations...

 

[phanbuey]

I need to run some benchmarks but i do find that 6200 memory context off, power down off, gear down off, CPPC to cache-aware.... bank hash (vs DIMM/rank) feels pretty amazing for stutter - stalker 2 feels like a real game

the 9800x3ds all kind of perform the same, which is kind of amazing - but some tweaks here and there do help (TREFI to 65535, high bandwith mode etc.)

 

[ShrimpBrime]

What about my question regarding load times? How can 96mb of cache matter when loading 8gb into ram at once?

It's obvious you need a better grasp on this.

Load times, is the time it takes for the SSD to write data to System RAM. This uses very little cpu cycles.

The game engine utilizes what's in the system memory and the CPU will utilize X amount of cores. Some games use the SSD for cut scenes, so loading times would be noticeable throughout game play.

Infinity Fabric and 3D Cache Bandwidth is so great, the memory frequency doesn't saturate the capabilities. In fact, you want the Infinity fabric to be as fast as possible. But it hits a wall. Then you must use GDM 1:2 ratio. Which is fine if your north of 8000mt/s. Thus performing similar to only 6000mt/s. In which each system can be ""tweaked"" for some minor performance gains.

____
So, set the XMP up, then manually input 6200 and see if GDM gets enabled. If not cool. If it does, measure the performance difference on your end to see if it's worth it.

 

[Visible Noise]

I guess I should go back to my 12700k and gtx 1070. You know, to save electricity.

I am in the habit of limiting my framerate through the Nvidia Control Panel to control the heat and noise in my system.

Thanks everyone for the input, ideas, and data. That ram comparison for 9800X3D really shows how small the differences are for this CPU.

You have a post history you know. You’ve already been told this.

 

[JustBenching]

I'd like to see a study on 1% and .01% lows related to ram speed on the 9800X3D. It's hard to believe there is no benefit.

I would assume that faster data-rate should mean faster loading times, right? How can the 3D cache of 96mb can make much difference when you're loading, say, 8 gigs into ram all at once?

Ram speed by itself is useless. Increasing ram from 6000 to 6200 will do nothing on its own, you need to tune all the timings to see some actual benefits. You can get around ~15% more performance by ram tuning and PBO +200 on an x3d (assuming CPU is your limiting factor of course)

EDIT. You can run gear 1 up to 6400, 6600 if you feed it a crapload of voltage and are lucky. But most gains will come from memory tuning than just upping the mhz so its kinda pointless.

 

[freeagent]

Maybe I am trippin.. but I swear 8000C36 feels smoother in msfs than 6400C28 does.

 

[Pets4Ever]

You have a post history you know. You’ve already been told this.

The focus before was about whether or not to purchase different ram, not adjust settings for the ram I already had. One costs money, the other costs time and a little electricity. I'm muting you, I have no use for people who feel the need to pore over post histories so they can post a "gotcha." Get a life

Ram speed by itself is useless. Increasing ram from 6000 to 6200 will do nothing on its own, you need to tune all the timings to see some actual benefits. You can get around ~15% more performance by ram tuning and PBO +200 on an x3d (assuming CPU is your limiting factor of course)

EDIT. You can run gear 1 up to 6400, 6600 if you feed it a crapload of voltage and are lucky. But most gains will come from memory tuning than just upping the mhz so its kinda pointless.

I need to read up on ram tuning.

There has to be at least one edge case where having slightly higher speed would be good?

Maybe I am trippin.. but I swear 8000C36 feels smoother in msfs than 6400C28 does.

I read weird things about MCFS. Some people swear it runs with fewer hitches with 64gb memory despite it technically being able to run on 16gb.

 

[Chrispy_]

Maybe I am trippin.. but I swear 8000C36 feels smoother in msfs than 6400C28 does.

They should be the same, 8000 CL36 is only around 2.8% slower than 6400C28 in absolute latency, which is likely too small a margin to feel. Also, if MSFS is streaming data from RAM at more than the 52GB/s DDR5-6400 is providing with a with a 4070Ti then something is very very wrong - I know MSFS streams real satellite imagery for ground textures on the fly but those are coming into VRAM at a few MB/s, not GB/s!! Google Earth can handle that on a pathetic 8Mbit/s DSL connection LOL...

Maybe your CPU isn't 100% happy at 6400 Gear1 and it's digging into DDR5's single-bit error correction and that's why it feels a bit crap? Dumb experiment, but have you tried lowering your Gear1 speeds to something very stable (like 5600 CL36) and then testing MSFS? It'll probably perform 5% percent slower in 1% lows because 5600-CL36 is leaving some performance on the table, but that's the sort of difference you wouldn't FEEL, it's just a margin that should be measurable.

If you can feel a night and day smoothness difference between 6400C28 and 5600C36 then something is wrong at 6400 - likely your IMC can't handle 6400 Gear1.

 

[Pets4Ever]

It's obvious you need a better grasp on this.

Load times, is the time it takes for the SSD to write data to System RAM. This uses very little cpu cycles.

The game engine utilizes what's in the system memory and the CPU will utilize X amount of cores. Some games use the SSD for cut scenes, so loading times would be noticeable throughout game play.

Infinity Fabric and 3D Cache Bandwidth is so great, the memory frequency doesn't saturate the capabilities. In fact, you want the Infinity fabric to be as fast as possible. But it hits a wall. Then you must use GDM 1:2 ratio. Which is fine if your north of 8000mt/s. Thus performing similar to only 6000mt/s. In which each system can be ""tweaked"" for some minor performance gains.

____
So, set the XMP up, then manually input 6200 and see if GDM gets enabled. If not cool. If it does, measure the performance difference on your end to see if it's worth it.

OK, so it seems like I should have looked up a unit conversion before asking my question. Apparently 6000 mt/s = 48000 Mb/s? If that's correct then I see why the SSD would be the limiting factor for loading times...damn should've gotten PCIE 5.0 x 4

I'll try 6200 and see what happens, and also research ram tuning.

 

[phanbuey]

Maybe I am trippin.. but I swear 8000C36 feels smoother in msfs than 6400C28 does.

My chip wasn't fully stable at 1:1 6400 and would crash in POE2 6400 at stock XMP - but was stable and super smooth at 1:2 6600 2200FCLK (predictive) same timings - it should have been WAY slower, but really wasn't but was definitely smoother.

 

[Pets4Ever]

My chip wasn't fully stable at 1:1 6400 and would crash in POE2 6400 at stock XMP - but was stable and super smooth at 1:2 6600 2200FCLK (predictive) same timings - it should have been WAY slower, but really wasn't but was definitely smoother.

In-game smoothness?

 

[Frick]

Maybe I am trippin.. but I swear 8000C36 feels smoother in msfs than 6400C28 does.

I've seen people on this very forum claim AMD Bulldozer felt smoother in Windows than the Intel counterparts. It's all subjective man.